MikroTik MCP provides a bridge between AI assistants and MikroTik RouterOS devices. It allows AI assistants to interact with MikroTik routers through natural language requests, executing commands like managing VLANs, configuring firewall rules, handling DNS settings, and more.
Source Code: https://github.com/jeff-nasseri/mikrotik-mcp
Market Place: https://mcp.so/server/mikrotik
I’d like to know community members’ opinions about this integration and how we can improve it in the future. Since AI is growing, it could be good to support more networking tools in the AI integration. I hope you enjoy the integration and I hope to hear from you all.
Amazing!. Maybe i dont click in "Allow always’ button… but interfacing in natural languages is awesome.
Permissions are the weak link.
Jarvis es around the corner! 
Can this mcp server be connected to local llm and used in offline mode?
Hey @TheSwitch! ofc you can, for example you can install https://lmstudio.ai/ on your local and set the mcp on your local llm.
LLMs should stay as far from managing critical infrastructure as possible.
Not even the big vendors use LLMs for management. Monitoring, analyzing, and chatbotting? Sure. But making changes to infrastructure via a stochastic blackbox is a horrible idea.
Hey @ThrowMeAwayDaddy! MCP is just a tool, and I completely understand your concern. As of today, I wouldn't encourage anyone to use AI to manage critical infrastructure in an enterprise production environment without human supervision. That said, given the rapid growth of AI, I believe newer versions of LLMs will have the capacity in the coming years to reliably maintain more critical infrastructure.
The MikroTik MCP itself is not tied to any specific LLM, nor does it encourage network engineers to use one over another. It's simply a tool that provides three transport protocols (STDIO, SSE, and Streamable HTTP), allowing any LLM to communicate with MikroTik. Just like any other MCP, it's up to individual teams and companies to decide which LLM they prefer to integrate with.
Of course, LLMs do make mistakes, which is why they may not yet be the best choice for fully controlling critical infrastructure. But that statement may only hold true for the next few years or even months, given the remarkable pace at which LLMs are evolving.
One of the reasons I created this topic in the MikroTik forum is to share our thoughts together and explore how we can make this tool safer for current and future LLMs, so that MikroTik engineers can eventually use it to facilitate their work with AI.
If you spot a security issue, or have a required or even nice-to-have feature suggestion, the community would be happy to work on it. Feel free to create an issue at: Issues · jeff-nasseri/mikrotik-mcp · GitHub